Create app.py

app.py

@@ -0,0 +1,64 @@
+from transformers import DistilBertTokenizer, DistilBertForSequenceClassification
+from flask import Flask, request, jsonify
+import torch
+
+# Initialize Flask app
+app = Flask(__name__)
+
+# Load pre-trained DistilBERT model and tokenizer
+model_name = "distilbert-base-uncased"
+tokenizer = DistilBertTokenizer.from_pretrained(model_name)
+model = DistilBertForSequenceClassification.from_pretrained(model_name)
+
+# Ensure the model is in evaluation mode
+model.eval()
+
+# Define a function to predict score and risk
+def predict_deal_qualification(inputs):
+    # Prepare input text
+    input_text = f"{inputs['industry']} {inputs['stage']} {inputs['amount']} {inputs['lead_score']}"
+    
+    # Tokenize the input
+    tokens = tokenizer(input_text, return_tensors="pt", truncation=True, padding=True, max_length=128)
+
+    # Run model inference
+    with torch.no_grad():
+        outputs = model(**tokens)
+
+    # Get prediction logits
+    logits = outputs.logits
+    score = torch.sigmoid(logits).item()  # Convert logits to probability (0-1 range)
+    
+    # Risk classification (for simplicity, let's map logits to Low, Medium, High)
+    risk_classes = ["Low", "Medium", "High"]
+    risk = risk_classes[torch.argmax(logits).item()]
+
+    # Dummy recommendation (customize this as needed)
+    recommendation = "Schedule another meeting before sending proposal."
+
+    return {
+        "score": round(score * 100, 2),  # Scale the score to 0-100
+        "confidence": round(torch.max(torch.softmax(logits, dim=-1)).item(), 2),
+        "risk": risk,
+        "recommendation": recommendation
+    }
+
+# Define an endpoint for deal qualification prediction
+@app.route('/predict', methods=['POST'])
+def predict():
+    # Get input JSON data from POST request
+    data = request.get_json()
+
+    # Validate input structure
+    if not all(key in data for key in ['industry', 'stage', 'amount', 'lead_score']):
+        return jsonify({"error": "Missing required input data"}), 400
+
+    # Predict using the pre-trained model
+    result = predict_deal_qualification(data)
+
+    # Return the prediction result as JSON
+    return jsonify(result)
+
+# Run the app
+if __name__ == '__main__':
+    app.run(debug=True, host="0.0.0.0", port=5000)